} ;'J \ .
~�q
,
.,
SDF74

nl! "", ":"1'IIl'j!
, '.~;
~''''
f''' ,'~'. ~
""'1 f/?' 1"1
: :
(:: r~
;

.." ~"i"'';'
'!

WW-l
.11 VItt"" �� �
""
DOMESTIC FINANCE STUDIES NO ltrf ::;. 4.

ON STRUCTURAL CHANGE A.~D DEBT SERVICING CAPACITY

Homi Kharas

rhe views presented in this paper are solely those of the author
d" not necessarily reflect the official opinions of the World Bank
Ilrt.:i
or ita affiliates.

August 1981

JOINT BANK-FUND LIBRARY

. Public and Private Finance Division JLC014322
De~elopment Economics Department
De',el ;)pmen t Policy Staf f
1

I. Introduction

It seems to have become conventional wisdom to claim that LDC debt

rescheduling problems are reflections of the failure of borrowers to invest

the:.r loans in appropriate productive activities)! In a macroeconomic

cont:ext, this is interpreted as the proposition that a low average propensity

to :.nvest out of foreign capital inflows is cause for concern. The prevalence

of t:his view appears rooted in two areas. First, it resembles the

macroeconomic analogue to the intuitive micro hypothesis that profit

max:.mizing firms will borrow as long as the marginal returns to investment

exct!ed the marginal costs. If a firm failed to invest its borrowed funds, it

wou:.d quic.kly go bankrupt. Second, it reflects the intertemporal coincidence

of t:wo substantive recent developments in international capital movements:

the rapid growth in volume of total outstanding debt and debt service

obLgations (and the concomitant growth in concern about defaults) and the

reduction in the earmarking of aggregate foreign loans for investment

pro.iects, accompanied by a dramatic change in the composition of the source of

funes towards commercial lenders.

A primary aim of this paper is to evaluate the operational relevance

of a.n approach which focuses on the use to which externally borrowed funds are

put I first present in Section II a general long-run theory of

crecitworthiness that suggests that present concerns over the degree to which

1/ "International financing must serve to increase productive investment in
debtor countries and to improve their capacity to repay their external
debt." Excerpt of speech by J. de Larosiere. ~naging Director of the IX:.
delivered before the Economic and Social Council ,f the L~, Geneva, July
3, L981.
2

countries invest externally borrowed money are misplaced. Other structural

parameters of the macroeconomy, particularly the marginal propensity to save

and the average capital/output ratio are much more important. In Section III,

I sllow that several stylized facts which have characterized recent historical

defc.ult cases can be readily explained within this theoretical framework.

Section IV then applies the theory to three countries, Brazil, Korea and Peru

which were chosen because of their diverse international credit experiences.

I f:,nd that defaults are associated with specific kinds of structural change

in rhe economy and that such changes are normally leading indicators of

defilUlt. These findings suggest that the methodology may be refined and

devE!loped into a short-run predictive model of defaults with obvious policy

imp:, ica t ions for both borrowers and lenders.

II. A ~del of Creditworthiness

Although defaults may be associated with short-run liquidity

pro")lems, one may look for the cause of defaulting in the context of more

long-run considerations related to the simultaneous accumulation of domestic

cap~tal and external debt. This focus has as a central premise the notion

tha: there exist a variety of sources, including IMF arrangements, reserve

run�-downs, short-term supplier credits and longer-term commercial eurocredits,

from which countries can readily obtain funds to tide over temporary liquidity

pro':)lems, if their long-run position is fundamentally sound 11 Thus, for

1/ For example Heller [1966] shows that the probability of default may be
independent of the variance of the current account if this variance is
known and countries adjust optimal ~eserve holdings in. response. Bilson
and Frenkel [1979] provide econometric evidence for a broad cross-section
of LDCs showing that reserve holdings are indeed positively related to
current account fluctuations.
3

countries with access to capital markets, the cause of both short-term cash-

flow problems and defaults is rooted in the long-run growth-cum-debt process.

The starting point of our long-run model is the general observation

that central governments and semi-autonomous public agencies are the main

borrcwers. Because of their public nature, investments by these agents are

ofter. not accompanied by cost-recovery programmes. Rather, the returns to the

inve~tments accrue to the private sector, while the government relies on

incrEases in the tax base (national output) to fuel its revenue requirements

for cebt servicing purposes. We assume that because of weaknesses in the

inst;tutiona1 framework, the government is constrained from arbitrarily

raisj.ng effective tax rates)! This wedge between the agents that benefit

(the private sector) and the agents that bear the repayment obligations (the

publ:.c sector) is the central externality that distinguishes this problem from

the (;ne in which a traditional treatment of optimal foreign borrowing yields

the J'u1e that the marginal rate of return to investment should be equated with

the T~rgina1 cost of funds~ We emphasize the importance of an appropriate

valuation of the marginal benefit of borrowing, in the face of constraints on

government macro-management.

We start by relating output to the domestic capital stock by a

simp:.e fixed-coefficients function.

3/ :~ome such constraints must also be present whenever income distributional
',eights are applied in project evaluation as in, for example, Little and
>tl.rrless [1974].

4/ .>ee, for example, Bardhan [1967].
4

Y = bK (1)

Aggregate private consumption is based on net-of-tax private income. As the

pr:~vate sector obtains the returns from all domestic capital,

( 2)

Tht~ government obtains revenue from taxes and from borrowing externally, F.

It first services its outstanding external debt, D, paying interest at a rate

r, and amortization at a rate, 0, and then allocates a fixed proportion of

re'llaining resources to investments.

(3)

As the private sector is not a borrower 2i, total investment in the country is

thtc. sum of government investment and local private savings.

I = IG + Sp = aK + a2 (F - (r+e)D) - Co (4)

b[(I-c) (l-t) + a2t]

Tht' change in outstanding debt over time is given as an accounting identity by

gross new borrowings less repayment of principal.

D = F - eD (5)

2/ The essence of the analysis remains unchanged if we allow private foreign
borrowing.
5

If the capital stock depreciates at a constant rate, 6, the full growth-cum

debt development may be written as a function of the level of capital inflows

and the initial values of domestic capital and outstanding debt.

(6 )

whel'e - 6.

Let the government choose foreign capital inflows in each time

per:.od to maximize some national income or consumption objective, subject to

the constraint that long-run creditworthiness be maintained, in the sense that

tax revenues are always expanding fast enough to cover interest on debt. This

ens11res that ever increasing new borrowings would not be required for debt

ser.rice payments. We show below that this constraint implies that borrowing

onlf takes place when a1 > r. The intuition is that a 1 reflects the marginal

fut .ire expansion in the tax base (and tax revenues) when one unit of domestic

capLtal is added to the economy, while r is simply the marginal future cost of

debt.

The optimal level of foreign borrowing at any time is determined by

setting the marginal benefit equal to the marginal cost. It can be shown that

in ,1 linear system as described above, these will be independent of the actual

lev~l of the state variables K and Dj} In a long-run framework, the terms on

which debt is contracted reflect the viability of the whole joint accunulaticn

~ The complete solution to the optimalcontrol problem is treated in more
depth in Kharas [1980J

patt of domestic capital and external debt. As this path is known (from (5)

and (6� once the borrowing profile is specified, there is no new information

contained in observations of the state variables as time progresses. Thus,

the shape of the supply curve facing the country remains constant; its

pos:.tion, howeve, is clearly subject to a scale factor. As world capital

marLets expand, supplier portfolio considerations imply a larger level of

bon'owing is available at any given interest rate, and so the amount of

bor,:owings that maintains marginal benefit equal to marginal cost expands at

thi:> exogenous rate, f.

F = Fo eft o <f <r (7)

The growth rate of new borrowings is constrained to be less than the interest

rat,~ to guarantee that the country does become a net supplier of real

res)urces to the rest of the world at some future time, a condition which

see�rlS necessary for the existence of a rational market solution}J

The solution to the system (5) - (7) is described by:

- a,

where the X's are the eigenvalues and the V's the associated eigenvectors.

As the two roots are real and of opposite sign, the system must be

characterized by a saddle point. Figure 1 below illustrates characteristic

paths for capital and debt accumulation. The scale along the axes ~ill be

expa.nding at a rate f over time.

2! This problem was first noted by Domar [1950]. To ensure a finite optimal
level of borrowing,it is clearly necessary to stipulate that real
resources do not come into the country at all times.
K

-_... \�1
2.

Figure 1

We define a creditworthy country, in this model, as one with the

ploperty of being on a path that leads to an ever-expanding capital stock. In

trms of Figure 1, if the initial position is above the eigenvector V2 , then

it will never be necessary continuously to run-down (or sell off) domestic

cl:.pital to meet debt-service obligations. The product from domestic capital

wi.ll be more than sufficient to cover desired consumption expenditure, capital

r:placement and interest on debt. Because the country remains creditworthy in

tbis sense, there should be no danger of external resources drying up. Debt

w:.ll be constantly rolled over. Both the contribution of foreign capital and

domestic output are, therefore, important in analyzing creditworthiness.

It is clear that the structure of the time paths along which

~Imestic capital and external debt progress will depend on the initial values

0:' these two variables relative to the contours shaped by the technical and

~!havioral parameters governing the position and slope of the eigenvectors of

the phase diagram. This joint variable/parameter relationship underlying
8

creditworthiness has often been overlooked, with emphasis almost exclusively

directed at the variables. The most common position is that if the values of

the variables appear inconsistent with a successful outcome, then debt

fin~nc1ng should be delayed. Failure to look at the parameters has made

ru1es-of-thumb for analyzing debt-servicing capacity in terms of indicators

hard to implement across different countries ��

From the eigenvector V2 in Figure 1, we get the trade-off between

the initial values of domestic capital and outstanding debt that leaves the

cocntry in a creditworthy situation. That is, if a country is on the margin

of creditworthiness (on V2 ) and then accumulates an extra unit of debt by

borrowing from abroad, then the domestic capital stock must increase by at

1e~st a2(r+e)/(a 1+e) if the country is to remain on or above V2 � But we know

from (6) that the actual increase in domestic capital due to one unit of

fo::eign borrowing is a2' Hence the constraint that a country can only borrow

if it maintains its creditworthiness is equivalent to the condition

a 2 ) a 2 (r+e)/(a 1+e) which simplifies to a 1 ) r. By mathematical induction,

we can then claim that this is a general criterion for successful borrowing.

Several important implications may be derived directly from this

bocrowing rule. First, there are a variety of factors embodied in the

parameter a 1 that influence creditworthiness. The proportion of foreign

borrowing that goes to investment is only relevant insofar as it is correlated

with the propensity to invest out of tax revenues. The efficiency of

investment (summarized in the capital-output ratio) is a key variable; but

perhaps more interesting is the presence of behavioural savings and tax

param~ters. Even if the marginal social return on investment is greater than

tre cost of externally borrowed funds, a government which invests in social

irfrastructure projects may run into debt-servicing problems unless the
9

aggregate marginal propensity to save is sufficiently high. This raises the

possibility of moral hazard problems whereby the act of borrowing induces

behc.vioural changes which reduce credl tworthiness � For example, i f public

invt!stments via subsidized state enterprises reduce opportunities for private

secl:or involvement, then unless perfect local capital markets exist, private

sav~ngs would fall J! Alternatively, access to foreign financing may reduce

the government's determination to collect tax revenues, again resulting in a

fall in aggregate domestic savings. These are cases where poor macro

manlgement creates problems by permitting a 1 to fall as borrowing takes

pla~e. A second implication, however, is that if al falls ~elow I' because of

some structural change in the behavioural and technological coefficients

inc:)rporated in aI' then debt servicing problems can arise even i f the initial

borrowing programme was well conceived. This merely reflects the common

result that there are risks attached to the fact that borrowing is associated

wit':'} a variable stream of benefits while servicing obligations are fixed. The

suggestion is that debt service problems may be associated with changes in

ai' Figure 2 illustrates this. A country has been borrowing and reaches the

pojnt A, with the expectation of maintaining its creditworthiness as growth

takes place. A sudden shock, (perhaps a terms of trade deterioration which

could lower savings rates), reduces a1 to a level a
1
< r < a 1� Now the

ex:.sting position is on an unstable path and i t is likely that both borrowers

an,i lenders will seek to reduce rapidly new capital inflows, resulting in

liquidity shortages and default.

~ The relationship between private savings and investment opportunities in
the presence of imperfect capital markets is derived in detail in
Virmani [1981] �
10

\
\

\1
V

Figure 2

III. Stylized Facts of Defaults

The relevance of the model to the real world can be ascertained by

its ability to explain certain stylized features associated with defaults and

the subsequent renegotiation procedure. There have, to date, been few

attempts at reconciling generalizations observed in historical case-studies

with a theoretical model. We examine briefly below the way in which the model

framework accounts for three stylized facts: the preference on the part of

lenders to reschedule debt on easier terms rather than writing-off existing

loans; the lack of general support by LDCs for generalized debt relief and

finally, the attempts made by creditors to ensure that aid is channelled

towards basic 'development' needs rather than simply credited against debt

ser'lice obligations.J.!

~ This does not imply that economic iactors are the only motivating factors,
but that the underlying economic interests may reinforce political
opinions.
11

During a rescheduling, creditors will typically agree to accept an

alternative stream of debt service payments with a present value substantially

bele,w that of the original obligations due J!l/ We may, therefore,

characterize the refinancing operations involved in defaults as being

equ:,valent to a simple unrequited transfer in perpetuity from the creditors to

the borrower. The amount of assistance, A, will depend on the country's

abLity to pay W As defined in our model and illustrated in Figure 2, the

purpose of the transfer would be to lower the eigenvector

V;, !;uch that given the actual level of domestic capital stock and outstanding

deb': the country's creditworthiness would be restored. Let the negotiation

pro,~ess *
result in a transfer of A. By calculating the full expression for

v; it is easily seen that such a transfer to the government will result in a

The present value of the loss to the
,
crelitors is simply A* /r. We now ask what quantity of debt, DR' would need to

be yritten-off by the creditors to yield the same gain in creditworthiness as

*
tha: provided by A. The write-off of outstanding debt is simply a leftwards

shirt of the initial point, by an amount inversely proportional to the slope

of il 2� DR is readily computed as A*(ai+9)/ai(r+6), and represents directly

the cost of the rescheduling exercise to the creditors. Comparing the

maglitude of losses suffered by the creditors under the two alternative

refinancing operations, the creditors would suffer a higher loss by rescinding

existing debt when ai < r. But this is precisely the condition which resulted

10/ Extended grace periods and lower interest are used to effect the
transfer. Of course, if only amortization is rescheduled at market
interest rates, then there is no assistance. But then we ask why a
rescheduling was necessary, rather than the country simply borrowing on
inter~ational markets. See for example, IMF [1981).

ll! ~e prinCiple of ability to pay in debt renegotiations is documented in
Cizauskas [1979).
12

in a default in the first place. Hence, we may claim that when a country

de::aults, creditors who desire to restore the country to a creditworthy

po:;ition at minimum cost to themselves would choose a refinancing operation

ra:her than writing-off a portion of their loans.

The symmetrical argument also indicates why LDCs have been unable to

ge1erate cohesive support in their bargaining for generalized debt relief in

th~ context of the North-South dialogueJl! For creditworthy countries, a 1 is

gr=ater than r, so they get more benefit from additional transfers than from

across-the-board reductions in debt where these yield the same improvement in

creditworthiness. Consequently, the focus of their efforts is in enlarging

the net transfer of additional resources.

Bitterman [1973] summarizes his analysis of debt refunding methods

by noting that aid has rarely been specifically directed towards providing

resources for debt service payments, being used instead to encourage desirable

development activities. We may paraphrase this in terms of the model to say

ttat donors attempt to influence domestic actions to raise a 1 to provide

ttemselves with an additional cushion against defaults. Similarly,

rEnegotiations are often linked to stabilization programmes designed to

irtprove public savings. Recipient countries, on the other hand, who are

originally presumably choosing al to maximize some intertemporal objectives,

r!!sist being forced to raise a l , and correctly note that the tying of aid in

this manner reduces the net benefit to the country. There is a difference

b!~tween the borrower's intertemporal trade-off and the lenders' desire to

~lximize creditworthiness. We may conclude this section by noting that the

c~editworthiness constraints yields economic arguments in support of the

lU See Roths tein [1979].
13

vartous positions defended by aid donors and recipients in historical

refunding exercises.

IV. Empirical application

The hypothesis that structural change in the economy, resulting in a

lowering in a1' is the major instigating factor behind defaults can be

emFirically investigated. Because a1 includes both behavioural and technical

parameters, it will reflect changes in the economic and sociopolitical

en"ironments. It has often been argued that in the last analysis it is the

lack of political confidence of foreign investors that triggers a default at

SOlle precise moment in time. One may imagine these foreign investors,

es)ecially the international commercial banks, deriving their information from

lo::al clients or branches of international clients. In these circumstances, a

de:~line in foreign confidence in the economy will always lag behind a decline

of local investor confidence. This latter would be reflected in a fall-off of

private savings and investment. The structural change considered is thus a

ctange i~ private intertemporal decisions. We investigate below the

association of changes in aggregate savings rates with default, for three

c(,untt'ies, Brazil, Korea and Peru which have had substantially different

e::periences in the international credit markets over the last 20 years.

T~ for Structural Change

The theory tells us both that downward changes in a1 will precede

d;~fault observations and that changes in a2 are likely to be irrelevant. We

d;) not wish, therefore, to restrict the test to the hypothesis that a sudden

change in a 1 at an ex ante determined time is significant in default events.

Father, we shall sacrifice the strength of a high-powered significance test of
14

an :.mposed formulation of the change in a 1 ll, and instead permit the data to
bring out any evidence of structural change by graphical methods. The

pro,~edure used was introduced by Brown, Durbin and Evans (1975].

A basic regression model is conditional on normal uncorrelated

err:>rs. If the specification of parameters as being constant over time is

inc)rrect, then we would expect this to show up in the behavior of the

residuals. It has been found that plotting the ordinary least-squares

residuals (or their squares) against time does not yield sensitive indications

of gradual changes in the parameters. In certain situations, the cumulative

sun or sum of squares is better; this latter, however, is not very amenable to

statistical tests of departure from the expected value. An alternative
ap~roach is to derive a new series of recursive residuals, W , defined as the
t
difference between an observation of the dependent variable Yt and the least

sq\,ares predictor of the observation, yt' based on a regress ion of y on the

incependent variables, X, over the period t = 1, ��� ,t-I. The recursive

ref:iduals are, therefore, based on changes in the estimates of the regression

pal:ameter vector, resulting from the consecutive addition of observations to
2
thl; sample. ..
It can be shown that the W. are N(O, cr) under the null

hypothesis that no structural change occurs. We expect that when a change

d04~s occur, then this will show up relatively quickly in the behavior of the

re�~ursive residuals.

Under the null hypothesis, the cumulative sum of the squared

re:ursive residuals, normalized by the cusum-squared statistic for the whole

salple, has a beta distribution with an expected value of zero at t :: k and 1

at t =T where k is the number of parameters to be estimated and T is the full
14
sanple size. , We may draw a confidence band around the expected value line,

131 Such as an exact Chow-test.

14/ cf. Brown, Derbin and ~vans [1975]

such that if the CU8um-squares series crosses outside the band, we may reject

the ~ull hypothesis at any required level of significance. The procedure may

be r~peated with residuals obtained by running the regression backwards over

time as well as forwards. This yields added information on changes which may

have occurred close to the starting point.

Brown etal. "prefer to regard the lines constructed in this way as

yardsticks against which to assess the observed sample path rather than

pro�dding formal tests of significance." An independent test, which can

further help locate departures from the null hypothesis is the Quandt log-

likelihood ratio test. This models an alternative hypothesis that the

pare,meters are constant until a given point in time when they switch to a

different, but still constant, value. If the alternative hypothesis is true,

then the log-likelihood ratio (LLR) will have a pronounced minimum at the time

of ':he switch. Although no formal test can be constructed for the min LLR, as

its distribution under the null hypothesis is not known, the plot of the LLR

indi.cates whether a sharp transition (marked, jagged minimum) or a gradual

cha'lge in the coefficients has taken place.

Res llts

The backwards and forwards cusum-squares and the Quandt log-

likelihood ratio tests were applied to the simple investment equation (6) for

Brazil, Korea and Peru over the period 1957-78. All the variables were

deflated by a price index. GDP was used as a proxy for the domestic capital

stc,::k. Although export taxes are a primary source of government revenue for

Peru, because exports are highly collinear with GDP, the GDP variable was

retained to facilitate comparison with the other results. Use of GDP implies

an emphasis on changes in savings rates as the main source of variation of the
paramet:er al.
16

Figure 3 shows the Brazilian cusum-squared tests. A 90% confidence

inter'lal is drawn, such that the null hypothesis that no structural change has

taken place is only rejected outside this interval. Clearly for any

predictive model of defaults the confidence interval would be adjusted

depending on the loss functions associated with Type 1 and Type 2 errors. If

failcre to predict defaults is more costly then wrongly predicting defaults,

then the confidence interval would be much narrower (perhaps only 20% or so)

than drawn here. For the present, as we are trying to establish the

relationship between structural change and defaults, the more stringent

interval is considered.

Both backwards and forwards cusum-squared plots show some evidence

of structural change around 1963/65, althoug~ this is not sufficiently marked

to rl!ject the null hypothesis. In addition, the backwards plot indicates

chan:~e in the 1966-68 period. Inspection of the numerical progression of the

coef:'icients as the length of the estimation period is increased (Table 1)

sugg~sts that the first change is associated with a rapid decline in a 1 which

~ott)med out in 1965 and then reversed itself into a sustained rise, which was

contLnued through the mid-1970's. By contrast, the movement in the a2

coefF.icient seems much less important, as well as being quite random over the

sample period. In fact, the confidence level at which a2 is significantly

diffl!rent from zero is quite low until the sample size reaches the full 22
observations.

Additional evidence that the pre-1965 investment regime was

significantly different from the post-1965 regime is provided by the Quandt

log-likelihood ratio plot which displays a deep minimum in 1965. To gain more

infermation on which coefficients were changing, the sample was partitioned

inte two at 1965 and separate regressions run for the two parts. The serial
17

F'i gure 3

Forwards and Backwards Cusum-Squared Plots

(1957-78) - Brazil

�
�
"
"
1960
"" \

~i~nificance line
. 965

""
"
1970 "
"
\
"
" "
" "
1975
....
'
"" " \
...
" "
" ""
1980 "
0.00 1.00

/ /
/ /
/ I
1960 / /
'O~ Sb:nificance line / /
/

I
/
1965 /
/
/
/ /
/ I
1970 / /
/ /
� /
/
I /
/ /
1975 / /

/ /

/ /
- 18

Table 1
Changes in Coefficients Over Sample Period - Brazil

for regression!' - ,.
K
p
t +
cons.
GDP
al p t a2
Net cap. inflows + e
p

CONSTANT t-Stat. a t-Stat. a2 , t-Stat.
l
1957 0.000 0.000 0.000 0.000 0.000 0.000
1958 0.000 0.000 0.000 0.000 0.000 0.000
1959 0.000 0.000 0.000 0.000 0.000 0.000
1960 0.000 0.000 0.000 0.000 0.000 0.000
1961 -0.218 - 0.682 0.237 1. 955 1.626 0.578
1962 -0.132 -0.635 0.210 2.410 0.800 0.472
1963 0.041 0.449 0.134 4.399 1.927 1.649
1964 0.035 0.422 0.140 5.442 1. 433 2.174
1965 0.119 0.849 0.126 2.855 -0.406 -0.566
1966 0.046 0.225 0.153 3.063 -0.410 -0.478
1967 -0.047 - 0.313 0.185 4.106 -0.054 -0.064
1968 -0.195 -1.598 0.232 6.656 0.623 0.815
1969 -0.274 - 2 . 503 0.257 8.587 0.794 1.021
B7C -0.307 - 3. 327 0.268 11 .059 0.940 1.307
1971 -0.300 -3.943 0.266 13.727 0.876 1.603
1972 -0.290 -4.325 0.263 15.750 0.806 1.650
1973 -0.271 -4.740 0.258 18.599 0.772 1.632
1974 -0.281 - 4. 861 0.262 19.237 0.230 1.111
1975 -0.294 -5.252 0.266 20.137 0.268 1 .321
1976 - 0.270 -5.379 . ~.O.259 22.671 0.316 1.606
1'377 -0.223 -5.074 0.246 27.389 0.493 2.813
197E -0.202 . -4.999 . 0.241 31. 06~ 0.555 3.306

2.' All variables drawn from International Financial Statistics
of the IMF.

Figure 4

Quandt Log-likelihood Ratio - Brazil

1960

1965

197C

197: -.--------______~~~--------------------~~~-------------
-20.00 -10.00
- 19

correlation, which is evident for the run using the whole sample and which

suggests that the constant coefficients model is misspecified, disappears. To

guard against the possibility that one or two bad data points, in the chaotic

years 1963-65 were dominating the regression results, a run was tried omitting

th '!se years; serial correlation remains. The evidence is very strong that

po~t-1965 Brazilian investment reflected a structural change in the economy,

with marginal savings rates sharply higher than before.

Table 2

Marginal Propensities to Invest Over Time - Brazil

1957-78

Period 1957-78 (omit.63-65) 1957-64 1966-78

Constant -202.72 -170.85 35.83 -78.69
(-4.999) (-5.481) (0.422) (-1.55)

GDP .2414 .2400 .1410 .2287
(31.06) (41.62) (5.44 ) (30.32)

Cap. inflows .5549 .4896 1.433 .5476
(3.306 ) (3.80) (2.17) (4.235)
-2
R .990 .9948 .8007 .9933
D.to] � 1.1577 1.3710 1.756 2.076

The credit history of Brazil reveals a succession of debt

consolidation and rescheduling exercises in 1953, 1955, 1958, 1961 and 1963-5

followed by a total absence of such problems up to the present. This is

clearly consistent with our theoretical framework suggesting that
- 20

creditworthiness is inherently tied to the magnitude of the parameter a 1 � The

extremely low value pre-1965 is associated with the periodic reschedulings,

but after measures were effected which substantially raised the marginal

propensity to invest out of domestic output the country was able to emerge

::rom this debt problem regime into one characterized by long-run

t:reditworthiness. The results are made more interesting by the coincidental

=all of the propensity to invest out of new capital inflows in the later,

~reditworthy period.

Korea provides an example of a country which has managed to avoid

debt servicing problems in the recent past, despite having to adjust to major

changes in the level and terms of capital inflows. For example, in the years

1955-62 the level of grant aid provided at least one-half of the total

resources available for capital accumulation, and in some instances much

more. After 1962, this gradually tapered off. The decline in the relative

importance of grants and concessional loans is even more clearly seen in the

proportion of imports covered by these inflows. On average, between 1953-62

this was 69%; in 1963-64, 41%; in 1965-69, 23% and in 1970-74, 11%~ In

conjunction with the authorities' desire to raise total capital inflows to

finance more rapid growth, the fall in concessional flows implied a rise in

the effective cost of foreign borrowing. This did not, however, lead to debt

servicing problems because the government of Korea introduced a new domestic

interest rate regime that stimulated private savings, and took measures to

raise public savings dramatically from 0.5% of GDP in 1964 to 2.8% in 1966.

Hence, the major financial reform of 1965 was in a direction which improved

creditworthiness sufficiently to outweigh the problems which may otherwise

have resulted from the rise in average foreign interest payments. The

12! See E. Mason etal. [1980], p. 190.
- 21

irrmediate impact of these measures was to encourage firms to borrow from

atroad and to create an environment in which the risks of default were low and

lenders would be prepared to lend. The theory predicts that the increase in

savings rates was probably a major factor in reducing risk and in opening up

a.:ceS9 of individual firms to international capital markets.

The rapid inflow into Korea of private, foreign capital post-1965

created problems of control over the monetary aggregates and a rise in

itlflation. The balance of payments, however, was kept in surplus by the

capital inflows. After attempts to limit monetary growth by restricting

commercial bank lending proved ineffective, the government elected to impose

tighter capital controls and banned certain types of loans, including those

Idth medium-term (3 year) pay-back periods. To maintain the investment

rlomentum, the authorities provided preferential credits for specified items

,md sectors. Both the structural changes in the economy mentioned above are

clearly seen in the forwards and backwards cusum-squared plots and in minima

i.n the Quandt log-likelihood ratio. Inspection of the change in the estimated

~oefficients of the investment function over time (Table 2) shows the dramatic

rise in savings out of domestic output from 0.091 for the 1957-64 period to

~.171 for 1957-66. Of course, the marginal increment in savings rates was

even greater than the increase in the period averages quoted above. The

second shift of parameters, after 1971, is again associated with a rise in the

savings rate, but accompanied by a fall in the propensity to invest out of

foreign capital inflows. This latter reflects the growing difficulty facing

private firms in obtaining foreign credits and the economic incentives they

faced to switch towards preferential domestic loans as a source of credit. As

expected, this decline has not led to significant debt problems for Korea; the

rise in savings rates more than offsets this.
22

Figure 5
Forwards and Backwards Cusum-Squared Plots - Korea

.
�
'\. '\.
'\.
'\.
'\.
'\.
1960
" '\.
'\.
'\.
1O~~ Silitnificance line

" '\.
'\.
1965
" '\.

"" '\.
1970
" '\.
'\.

"
1975 ' ....
-. ""
, '\.
,
0.00 1.00

/ /
/ /
/ /
i9O /
/
10,';, Significance line / /
/
/

/
1965 /
/
/
/
/
1970 /
/
/
/

/
/
/
1975 / /
/ /
/ /
- 23
Table 3

Changes in Coefficients OVer Sample Period - Korea
a/ K
for regression- p -
cons.-
t +
a1
GDP.j.
a2
Net cap. inflows +
-p-
p e

CONSTANT t-Stat. a
1
t-Stat. a
2 t-Stat.
1957 0.000 0.000 0.000 0.000 0.000 0.000
1958 0.000 0.000 0.000 0.000 0.000 0.000
1959 0.000 0.000 0.000 0.000 0.000 0.000
1960 0.000 0.000 0.000 0.000 0.000 0.000
1S61 -0.972 -1. 962 0.160 4.657 0.086 0.365
1962 -l. 426 - 3.541 0.168 4.450 0.294 1.474
1963 -1. 055 - 3.602 0.142 4.220 0.343 1.641
1964 -0.656 - 3.571 0.091 6.273 0.632 4.968
1965 -1. 049 -2.265 0.148 4.866 0.280 0.922
1966 -2.265 -2.869 a. 171 2.801 0.725 1.231
1967 -2.804 -4.137 0.171 2.707 1.020 1.835
1968 -2.971 -4.813 O. 155 2.701 1.265 3.004
1969 - 3.171 -5.689 0.160 2.841 1.312 3.199
1970 -3.476 --7.848 0.191 4.333 1 . 127 3.187
1971 - 3. 08::1 - 6. 712 0.164 3.406 1.232 3.085
1972 -3.312 - 8.300 0.209 10.936 0.871 4.717
1373 - 3. 533 -7.467 0.253 19.204 0.511 3.249
1974 -3.155 -7.675 0.252 18.481 �0.380 2.831
1975 - 3. 184 - 8.532 0.253 20.003 0.382 2.937
1976 -3.200 - 8.893 0.256 33.720 0.354 3.883
1977 - 3. 246 - 7 . 684 '. 0'.271 42.804 0.202 2.346
~/ All variables drawn from International Financial Statistics
.,E the IMF.

Figure 6

Quandt Log-likelihood Ratio Plot - Korea

1960

1965

1970

1975

�30.00 -20.00 -10.00

- 24

Peru has experienced continued reschedu1ings over the past twenty

years. the most recent being in 1979/80. Despite the fact that the country

has large collateral in the form of oil and other mineral reserves which would

~eem to augur well for long-term prospects, and which provides the base for

Peruvian participation in euromarkets on a large scale, creditors have

periodically lost confidence in the economic development path, forcing the

authorities to undertake stabilization measures and renegotiate debt service

obligations. 1963 was a year of formal reschedu1ings as was 1968. Default

was marginally averted in 1975 and in 1976 the government announced a new

series of stabilization measures which were to be monitored by a steering

committee of private banks. The breakdown of the programme culminated in a

broad restructuring of debt in December 1978.

The forwards cusum-squared plot of the investment function shows

signs of structural change in 1963 to 1968 and this is further brought out in

the backwards run. The Quandt log-likelihood ratio shows minima in 1963, 1967

and a steep decline into a trough in 1974. The first two declines in the LLR

correspond to crisis times which were precipitated by a fall in the marginal

propensity to invest out of domestic output. It is interesting to note that

although earlier rescheduling exercises included attempts to raise domestic

savings, these measures were ineffective. For example. in 1968 the government

tried to curtail expenditures, raise taxes and restrict credit, but as we can

see in Table 4, the marginal propensity to invest out of domestic output was

not raised. Similarly, the stabilization program of 1976/77 failed to improve
the situation.

It is of particular interest that in the Peruvian case the aftermath

of the formal defaults was ~ return to the old situation, in strict contrast

to the 1965 Brazilian experience. Examination of the changes in coefficients
- 25

s~own in Table 4 shows that although the propensity to invest out of net

capital inflows was increased from 0.286 for the period 1957-67 to 0.539 for

period 1957-77, the propensity to invest out of domestic output remained

Essentially unaltered at 0.076 and 0.075 for these sample periods

lespectively. Hence, there is no indication that the renegotiations of debt

over this period were sufficiently broad to restore long-run creditworthiness,

nor was Peru able to initiate programs under which future borrowing would help

~elieve the situation. Indeed, between 1970-75, the ratio of gross domestic

Lnvestment in GDP rose from 13% to 19%, mostly attributable to substantial

foreign capital inflows and steady rises in the propensity to invest out of

these inflows, but due to the low magnitude of the savings rate, this process

did not improve stability.
26 -

Fi gure 7
Forwards and Backwards Cusum-SQuared Plots
{1957-77} - Peru

I ", "
, "
" "
,
1960 I " 10~ Significance lines
"
!
,
" " ,
I "", " ,

1965
"
",
I ,
,,
1970 !
"
,
" "
",
. "
1975
....."
" "- ",
I " ""
I

0.000 " 1.000

/
/ /

1960, / .'
1O~ Significance lines / /
/
/
/
1965 /
/
/
/
/ /
1970 / /
/ /
/ /
/ /
/ /
1975 / /
1
/
I
/
/ /
- 27
Table 4

Change in Coefficients Over Sample Period* - Peru
.

K GD~
for Regression!! p � const. + a 1 p a2
Net cap. inflows +
p e

CONSTAnT t-Stat. a
l
t-Stat. a2 t-Stat.
1957 0.000 0.000 0.000 0.000 0.000 0.000
1958 0.000 0.000 0.000 0.000 0.000 0.000
1959 0.000 0.000 0.000 0.000 0.000 0.000
1960 0.000 0.000 0.000 0.000 0.000 0.000
1961 -0.494 -1. 570 0.387 3.010 1. 291 3.631
1962 -0.481 -3.211 0.381 6.467 1.279 5.665
1963 0.014 0.073 0.184 2.546 0.586 1.825
1964 0.079 0.491 0.156 2.610 0.619 2.031
1965 0.185 1.504 0.115 ' 2.598 0.471 1. 752
1966 0.219 2.431 0.102 3.277 0.424 1.832
1967 0.290 3.726 0.076 2.904 0.286 1.300
1968 0.316 4.077 0.064 2.547 0.376 1.753
1969 0.325 4.423 0.060 2.567 0.426 2.229
1970 0.319 4.578 0.063 2.865 0.392 2.312
1971 0.332 5.445 0.058 3.133 0.378 2.354
1972 0.328 6.120 0.060 3.741 0.383 2.520
1973 0.323 7.438 0.061 4.960 0.393 2.911
1974 0.301 8.644 0.068 7.168 0.479 5.277
1975 0.296 9.092 0.069 7.835 0.509 8.226
1976 0.282 8.384 0.074 8.172 0.540 8.523
1977 0.283 8.897 0.075 8.690 0.539 8.789

!of All variables drawn from International Financial Statistics
of the IMP.
* No accurate data available for 1978.

Figure 8

quandt Log-likelihood Ratio Plot -'Peru

1965

1970

1975

-15.0�J -10.00

- 28

There are three generalizations that emerge from the case studies.

First, the renegotiations are associated with a fall in the marginal

propensity to invest out of domestic output. By using sensitive techniques

for detecting structural change, useful information on the likelihood of a

default in the near future can be generated. Second, a successful

renegotiation, in the sense that a country emerges into a long-run

creditworthy regime, involves a substantial rise in the level of the savings

rate. Refinancing exercises that fail to generate more domestically-financed

investment are typically followed by further reschedulings. Similarly, if

increasing recourse to more expensive commercial loans is taken, maintenance

of creditworthiness necessitates an improved savings performance. The

Brazilian and Korean cases are examples of how countries with poor historical

records could have been predicted as having a low risk of future defaults on

debt servicing payments once they managed to alter the fundamental conditions

affecting creditworthiness. The Peruvian example, on the other hand,

illustrates the ability to predict continued debt-servicing problems even

after renegotiations had alleviated the most immediate pressures. Third, the

degree to which foreign capital inflows are invested and changes in this

coefficient are not especially important. We show examples where differences

between countries are inversely correlated with creditworthiness (Korea vs.

Peru), and where the movement of one country from an uncreditworthy to a

creditworthy environment is associated with a fall in the propensity to invest

out of external funds-(Brazil). A corollary of this is that attempts by an

inherently uncreditworthy country to emerge into a viable growth system by

borrowing heavily from abroad and investing the funds domestically, will

Simply exacerbate the future debt service problems unless savings rates are
bOosted.
- 29

Conclusion

We have developed a theoretical framework in which a country may

renE~ge on its debt-servicing obligations when payment of these in full would

sevE:rely cripple consumption and investment plans. The essence of the problem

is t:.hat the main borrower of external credits, the public sector, often allows

the returns on its investments to accrue directly to the private sector. As

genl~ral government collection of real resources to repay foreign debt depends

on~he overall growth of the economy, public and private savings playa

dom~nant role in the appropriate valuation of the marginal benefit of

bor:-owing and hence in determining creditworthiness. This framework was used

to ~xplain the economic underpinnings of the positions advanced by various

agelts (lenders, creditworthy and uncreditworthy borrowers) in the context of

the North-South dialogue. We suggest that the two considerations of

crelitworthiness and of resource transfers from North to South, would yield

the observed pattern of concerns. This is taken as justification of the

relevance of the theory to the real world.

The implication that creditworthiness depends on the marginal

propensity to invest out of domestic capital relative to the real cost of

external funds yields empirically testable hypotheses of a positive nature

(th,:it defaults are accompanied by a fall in savings rates) and a negative

nat'.lre (that the level of and changes in the propensity to invest out of

foreign capital inflows is unimportant). The results from three case studies

sUFPort these proportions both across countries and for one country over

tin.e. A technique for identifying structural change in the investment

eqtation by the use of recursive residuals is shown to be useful in predicting

WhEther a default is likely in the short-term future. The cusum-squared
- 30

analysis used grows out of a more general methodology designed to detect

changes in variable means in industrial quality control. It is much more

sensitive than an analysis based on actual residuals and is, therefore,

part.icularly suited to the problem of predicting potential debt crisis

sit\,ations for individual countries.
REFERENCES

Bardhan, P.K. 1967 "Optimum foreign borrowing," in K. Shell (ed.), Essays on
the Theory of Optimal Economic Growth M.I.T., Cambridge, Mass.

BilsOl, J.F. and J.A. Frenkel, 1979. "Dynamic Adjustment and the Demand for
International Reserves," National Bureau of Economic Research, Working
Paper No. 407, Cambridge, MA.

Bitterman, H.J., 1973. The Refunding of International Debt, Durham, N.C.
Duke University.

Brown, R.L., J. Durbin and J.M. Evans, 1975. "Techniques for Testing The
Constancy of Regression Relationships Over Time," Journal of the Royal
Statistical Society, B, 37, No.2.

Cizauskas, A.C., 1979. "International Debt Renegotiation: Lessons From
the Past," World Development, February.

Doma I, E. S " 1950. "Foreign loves tment and the Balance of Payments,"
American Economic Review, December.

HellEr, H.R., 1966, "Optimal International Reserves," Economic Journal
June.

Inter'national Monetary Fund .1981. "External Indebtedness of developing
count.ries", Occasional Paper.

Kharcls, H.J., 1980. "Optimal Foreign Borrowing by Less Developed Countries
Theory and Practice," unpublished Ph.D. dissertation, Harvard
University, Cambridge, Mass.

LitLe, ~.D. and J.A. Mirrless, 1974. Project Appraisal and Planning for
Developing Countries, Heinemann, London.

MaSO"l, E. etal. 1980. The Economic and Social Modernization of the Republic

of Korea, Harvard University Press, Cambridge, Mass.

MathLeson, D.J., 1979. "Financial Reform and Capital Flows in a Developing
Economy," IMF Staff Papers, September.

Puz, R., 1977. "How to Find Out When a Sovereign Borrower Slips From A-I
to C-3," Euromoney, December.

Rothstein, R.L. 1979. Global Bargaining. Princeton University Press,

Princeton, N.J.

Sargen, N., 1977. "Use of Economic Indicators and Country Risk Appraisal,"

Economic Review of the Federal Reserve Bank of San Francisco, Fall.

Solcmon, R.A., 1977. "A Perspective on the Debt of Developing Countries,"

Brookings Papers on Economic Activity, 2.

van Agtmael, A., 1976, "Evaluating the Risks of Lending to Developing
Countries," Euromoney, April.

Virmrlni, A., 1981. "The Nature of Credit Markets in Less Developed
Countries: A Framework for Policy.Analysis; Domestic Finance Study No.
71, World Bank, Washington, D.C.